On a late May afternoon in 2026, Argent Capital Management portfolio manager Jed Ellerbroek sat down with BNN Bloomberg and named three companies poised to ride the next wave of artificial intelligence demand. Among them was Microsoft—a name familiar to every Windows user, but now increasingly a bet on the economics of AI scarcity.

Ellerbroek's pitch, delivered on May 29, didn't lean on the software giant's legacy. Instead, it framed Microsoft as the ultimate infrastructure play for an era where AI compute is the new oil. Alongside semiconductor giant Broadcom and aerospace parts supplier TransDigm, Microsoft stood out as a cloud hyperscaler turning capital-intensive AI investments into recurring enterprise revenue.

His thesis: AI demand is creating a scarcity of compute, and the companies that control that compute—through silicon, networking, or cloud platforms—will capture outsized value. For Microsoft, it's the seamless integration from data center to desktop that makes the scarcity play so compelling.

Azure's AI Infrastructure Boom

Microsoft's Azure cloud platform has been on a tear, with AI workloads driving a significant portion of recent growth. By 2026, enterprises have moved beyond experimentation; they're deploying production AI models that require massive, sustained compute. This has pushed Azure's revenue growth above 30% for consecutive quarters, with AI services alone contributing an ever-larger slice.

Ellerbroek highlighted that scarcity isn't just about GPU supply—it's about who can offer enterprises a repeatable, secure, and integrated AI environment. Azure, with its global data center footprint and pre-built AI services, allows companies to focus on their own models without fretting about underlying infrastructure. That stickiness translates into long-term contracts and high switching costs.

Microsoft has not stood still. The company's multi-billion dollar infrastructure expansion, including new data center regions built specifically for AI, has positioned it to meet demand that far outstrips supply. The scarcity premium is real, and Azure is capitalizing on it.

Custom Silicon: The Maia and Cobalt Edge

One of the most underappreciated facets of Microsoft's AI strategy is its move into custom silicon. The Maia AI accelerator and Cobalt CPU, first announced in 2023, are now in full-scale production and powering Azure's most demanding AI workloads.

Instead of relying entirely on NVIDIA GPUs, Microsoft can offer customers a choice of compute architectures. Maia chips, designed for large language model training and inferencing, provide price-performance advantages that are increasingly appealing in a cost-conscious enterprise market. Meanwhile, Cobalt ARM-based processors handle general-purpose cloud tasks more efficiently than traditional x86 chips, reducing Microsoft's own power costs and improving margins.

Ellerbroek pointed out that custom silicon is a moat. By controlling the hardware and the software stack—from Azure OpenAI Service to Copilot—Microsoft can optimize the entire pipeline. It's a vertical integration strategy reminiscent of Apple's iPhone playbook, applied to the enterprise AI market.

Enterprise IT Procurement: A Microsoft Lock-in

Walk through any Fortune 500 IT department today, and you'll find Microsoft everywhere: Windows 11 on desktops, Microsoft 365 for productivity, Azure Active Directory for identity, and increasingly, Azure AI for building custom copilots. This ecosystem lock-in is a procurement officer's dream (or nightmare, depending on perspective) because it simplifies vendor management while making it very difficult to rip out any single component.

Ellerbroek emphasized that enterprise IT spending is consolidating around platforms that offer end-to-end solutions. Microsoft's ability to sell Copilot for Microsoft 365, which ties into Azure AI services, means each new AI feature pulls through more cloud consumption. The AI scarcity narrative thus extends to the desktop: as users demand AI assistance, IT departments must invest in the backend compute that powers it.

The Windows ecosystem plays a crucial, often overooked role. Windows 11's deeply integrated Copilot—available directly from the taskbar—drives usage that translates into Azure consumption. Every time a knowledge worker drafts a report in Word with AI assistance, it generates API calls to Azure. It's a virtuous cycle: more users, more AI, more cloud.

Windows as an AI Terminal

Microsoft's vision of the AI PC, first outlined years ago and now a reality in millions of devices, turns every Windows machine into a terminal for Azure AI. Features like live captions, camera effects, and background blur rely on local AI, but the heavy lifting—large model inferencing—still happens in the cloud. The integration between the two is seamless, thanks to the Windows Copilot Runtime and Azure AI hybrid infrastructure.

For enterprises, this means a predictable growth path for Azure spending as they roll out Windows 11 upgrades. Windows 365 Cloud PCs further blur the line between local and cloud, streaming a full desktop from Azure. In this model, the PC becomes a thin client to Microsoft's AI cloud, and every enterprise seat becomes a recurring revenue stream.

Ellerbroek didn't call out Windows by name, but the numbers make the connection obvious: increased Windows adoption—especially in AI-forward organizations—directly correlates with Azure AI revenue. That's an under-the-radar driver that Wall Street sometimes misses but that Argent Capital is betting on.

The Economics of AI Scarcity

The term "AI scarcity" gets thrown around a lot, but what does it actually mean for investors? Simply put, the demand for AI compute—measured in petaflops, GPU-years, or simply watts—exceeds supply by a factor of two to three times, according to industry estimates. Hyperscalers that have secured long-term energy contracts, chip allocations, and buildable land are in a prime position to meet that demand at a premium.

Microsoft fits that bill perfectly. It has committed to powering all its operations with clean energy by 2025 and has locked in deals with renewable providers for the coming decades. Its data center construction is ahead of schedule in key markets like Northern Virginia, Phoenix, and Dublin. When enterprise customers come knocking for AI capacity, Microsoft can actually say yes.

This supply-demand imbalance is turning into favorable unit economics. Azure's AI inference services, for example, command higher margins than traditional compute or storage. As more businesses shift from training to inferencing—where Microsoft has optimized its infrastructure—the revenue mix becomes even more profitable.

How Microsoft Compares to Broadcom and TransDigm

Ellerbroek's other picks—Broadcom and TransDigm—serve as interesting contrasts. Broadcom makes networking chips that are critical for connecting AI accelerators. It's a picks-and-shovels play on AI, similar to how Microsoft sells the shovels (cloud infrastructure) rather than just the gold (AI applications).

TransDigm, an aerospace parts manufacturer, might seem out of place, but it was pitched as a way to play the increasing complexity of defense and commercial aircraft—a market where digitalization and AI are also becoming mission-critical. Still, for pure-play AI exposure through an enterprise lens, Microsoft stands alone on that list.

The key difference: Microsoft has a direct line to the end user through Windows and Office. Broadcom sells components; Microsoft sells outcomes. This gives it pricing power and a feedback loop that Broadcom can't replicate.

Risks on the Horizon

No investment thesis is without risks. Microsoft's AI spending requires immense capital, and any slowdown in cloud growth could pressure margins. Competitors like Amazon Web Services and Google Cloud aren't standing still; they're investing heavily in their own AI infrastructure and custom silicon.

Regulatory scrutiny is another wildcard. As AI capabilities advance, governments worldwide are eyeing tighter controls on cloud providers. Data sovereignty laws could force Microsoft to build more local data centers, raising costs. Additionally, antitrust concerns—especially in Europe—could hamper its ability to bundle AI services with Windows and Office.

And then there's the valuation question. By mid-2026, Microsoft's stock trades at a premium multiple, reflecting high growth expectations. Any miss on Azure revenue could lead to sharp corrections. Ellerbroek acknowledged these risks but argued that the AI demand curve is still in its early innings, and Microsoft's integrated platform is the safest way to play it.

The Road Ahead

Looking forward, Microsoft's AI scarcity advantage is likely to persist for at least the next two to three years. The company's product pipeline—from the next-generation Windows Cloud PC to deeper Copilot integration across Dynamics and Power Platform—will only tighten its grip on enterprise IT budgets.

Investors who buy Microsoft today aren't just buying a software company; they're buying a bet on the infrastructure layer of the AI revolution. As Jed Ellerbroek put it on BNN Bloomberg, "In a world where AI compute is scarce, owning the platform that everyone needs is a pretty good place to be."